The Signal and the Noise: Why So Many Predictions Fail—But Some Don't

Very good and slightly surprising book, only missing out on 5* due to an understandable American slant.

Surprising, despite the sub title, as from a famous analyst/statistician this is more about how predictions fail, where the issues are, rather than how brilliant they are and how everything will soon be solved. As the author puts it less about what we know, than the difference between what we know and what we think we know.

The tone is exploratory, humble, and insightful. Much more by the way of understanding the world rather than explaining it - with a good dose of autobiographical detail thrown in too. Nate Silver is a very good writer, painting a vivid picture of the places and situations described, as well as explaining complex situations to general readers.

Most sections explore a topic in prediction or statistics through a lived experience or major world event. This gives the book a slightly uneven feel, some chapters are far more engaging than others, especially I feel for non-Americans. The weakest chapters were on baseball, elements of the climate change one, and terrorism. The book was far stronger on the financial crisis, weather systems, epidemics, chess, and Bayes theorem. Some others on poker and politics were middling.

He is able to distil excellent general points, more data does not mean better prediction, with ever more noise it may be harder to hear the signal. Exponential data increases as we have seen since the books publication leave us with more hypothesis and more false positives. Are we sure we were right with the prediction for the reasons we say, or was it something else? We have more data, but the same amount of truth as 1000 years ago. Proper analysis of long term predictions is rare.

The distinction on tactics vs strategy regarding chess and computer power was enlightening. Brilliant for its time at least, the developments in deep mind have made even these human domains fall now. Be interesting to see Nate's updated thoughts on the matter.

Insightful and prescient on the stock market when talking about the .com boom are crash - the parallels to today are astonishing, we are "buying a particular vision of tomorrow" unattached form realities - see Tesla stock price rise! (not that it will fall necessarily). Misaligned incentives for annalists create market irregularities and bubbles.

While part of the climate change section is outdated he is great when talking about scientist who don't forecast and forecasters who don't do the science, as well as a cutting analysis of both the political and media landscape regarding the issue. So far simple models have been better, yet he discounts any possible major shifts in the future as they have not/had not happened yet (such as methane release, rapid ice melting, tipping points tipping...)

Both the analysis and characterisation of 'us' (the United States) in the terrorism chapter are weak and telling. Clearly affected by 9/11 he seems very worried of another even bigger (nuclear or biological) attack in the US. This seems incredibly unlikely give the relative capabilities of any terrorist organisation today. And there is no talk of the cost or consequences of the US response to 9/11 - which have been worse for both 'us' and to a far larger extent anyone unfortunate enough to be 'them'. He does speak well of the Israeli response to terrorism, one of a criminal action not an existential threat. Though they have a aggressive stance on genuine existential threats - see Iran and nuclear development!